x86/cfi,bpf: Fix BPF JIT call

Peter Zijlstra · Alexei Starovoitov · commit 4f9087f16651 · 2023-12-15T16:25:55.000-08:00
The current BPF call convention is __nocfi, except when it calls !JIT things, then it calls regular C functions. It so happens that with FineIBT the __nocfi and C calling conventions are incompatible. Specifically __nocfi will call at func+0, while FineIBT will have endbr-poison there, which is not a valid indirect target. Causing #CP. Notably this only triggers on IBT enabled hardware, which is probably why this hasn't been reported (also, most people will have JIT on anyway). Implement proper CFI prologues for the BPF JIT codegen and drop __nocfi for x86. Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Link: https://lore.kernel.org/r/20231215092707.345270396@infradead.org Signed-off-by: Alexei Starovoitov <ast@kernel.org>
diff --git a/arch/x86/include/asm/cfi.h b/arch/x86/include/asm/cfi.h
@@ -9,15 +9,125 @@
  */
 #include <linux/bug.h>
 
+/*
+ * An overview of the various calling conventions...
+ *
+ * Traditional:
+ *
+ * foo:
+ *   ... code here ...
+ *   ret
+ *
+ * direct caller:
+ *   call foo
+ *
+ * indirect caller:
+ *   lea foo(%rip), %r11
+ *   ...
+ *   call *%r11
+ *
+ *
+ * IBT:
+ *
+ * foo:
+ *   endbr64
+ *   ... code here ...
+ *   ret
+ *
+ * direct caller:
+ *   call foo / call foo+4
+ *
+ * indirect caller:
+ *   lea foo(%rip), %r11
+ *   ...
+ *   call *%r11
+ *
+ *
+ * kCFI:
+ *
+ * __cfi_foo:
+ *   movl $0x12345678, %eax
+ *				# 11 nops when CONFIG_CALL_PADDING
+ * foo:
+ *   endbr64			# when IBT
+ *   ... code here ...
+ *   ret
+ *
+ * direct call:
+ *   call foo			# / call foo+4 when IBT
+ *
+ * indirect call:
+ *   lea foo(%rip), %r11
+ *   ...
+ *   movl $(-0x12345678), %r10d
+ *   addl -4(%r11), %r10d	# -15 when CONFIG_CALL_PADDING
+ *   jz   1f
+ *   ud2
+ * 1:call *%r11
+ *
+ *
+ * FineIBT (builds as kCFI + CALL_PADDING + IBT + RETPOLINE and runtime patches into):
+ *
+ * __cfi_foo:
+ *   endbr64
+ *   subl 0x12345678, %r10d
+ *   jz   foo
+ *   ud2
+ *   nop
+ * foo:
+ *   osp nop3			# was endbr64
+ *   ... code here ...
+ *   ret
+ *
+ * direct caller:
+ *   call foo / call foo+4
+ *
+ * indirect caller:
+ *   lea foo(%rip), %r11
+ *   ...
+ *   movl $0x12345678, %r10d
+ *   subl $16, %r11
+ *   nop4
+ *   call *%r11
+ *
+ */
+enum cfi_mode {
+	CFI_DEFAULT,	/* FineIBT if hardware has IBT, otherwise kCFI */
+	CFI_OFF,	/* Taditional / IBT depending on .config */
+	CFI_KCFI,	/* Optionally CALL_PADDING, IBT, RETPOLINE */
+	CFI_FINEIBT,	/* see arch/x86/kernel/alternative.c */
+};
+
+extern enum cfi_mode cfi_mode;
+
 struct pt_regs;
 
 #ifdef CONFIG_CFI_CLANG
 enum bug_trap_type handle_cfi_failure(struct pt_regs *regs);
+#define __bpfcall
+extern u32 cfi_bpf_hash;
+
+static inline int cfi_get_offset(void)
+{
+	switch (cfi_mode) {
+	case CFI_FINEIBT:
+		return 16;
+	case CFI_KCFI:
+		if (IS_ENABLED(CONFIG_CALL_PADDING))
+			return 16;
+		return 5;
+	default:
+		return 0;
+	}
+}
+#define cfi_get_offset cfi_get_offset
+
 #else
 static inline enum bug_trap_type handle_cfi_failure(struct pt_regs *regs)
 {
 	return BUG_TRAP_TYPE_NONE;
 }
+#define cfi_bpf_hash 0U
 #endif /* CONFIG_CFI_CLANG */
 
 #endif /* _ASM_X86_CFI_H */
diff --git a/arch/x86/kernel/alternative.c b/arch/x86/kernel/alternative.c
@@ -30,6 +30,7 @@
 #include <asm/fixmap.h>
 #include <asm/paravirt.h>
 #include <asm/asm-prototypes.h>
+#include <asm/cfi.h>
 
 int __read_mostly alternatives_patched;
 
@@ -832,15 +833,43 @@ void __init_or_module apply_seal_endbr(s32 *start, s32 *end) { }
 #endif /* CONFIG_X86_KERNEL_IBT */
 
 #ifdef CONFIG_FINEIBT
+#define __CFI_DEFAULT	CFI_DEFAULT
+#elif defined(CONFIG_CFI_CLANG)
+#define __CFI_DEFAULT	CFI_KCFI
+#else
+#define __CFI_DEFAULT	CFI_OFF
+#endif
 
-enum cfi_mode {
-	CFI_DEFAULT,
-	CFI_OFF,
-	CFI_KCFI,
-	CFI_FINEIBT,
-};
+enum cfi_mode cfi_mode __ro_after_init = __CFI_DEFAULT;
+
+#ifdef CONFIG_CFI_CLANG
+struct bpf_insn;
+
+/* Must match bpf_func_t / DEFINE_BPF_PROG_RUN() */
+extern unsigned int __bpf_prog_runX(const void *ctx,
+				    const struct bpf_insn *insn);
+
+/*
+ * Force a reference to the external symbol so the compiler generates
+ * __kcfi_typid.
+ */
+__ADDRESSABLE(__bpf_prog_runX);
+
+/* u32 __ro_after_init cfi_bpf_hash = __kcfi_typeid___bpf_prog_runX; */
+asm (
+"	.pushsection	.data..ro_after_init,\"aw\",@progbits	\n"
+"	.type	cfi_bpf_hash,@object				\n"
+"	.globl	cfi_bpf_hash					\n"
+"	.p2align	2, 0x0					\n"
+"cfi_bpf_hash:							\n"
+"	.long	__kcfi_typeid___bpf_prog_runX			\n"
+"	.size	cfi_bpf_hash, 4					\n"
+"	.popsection						\n"
+);
+#endif
+
+#ifdef CONFIG_FINEIBT
 
-static enum cfi_mode cfi_mode __ro_after_init = CFI_DEFAULT;
 static bool cfi_rand __ro_after_init = true;
 static u32  cfi_seed __ro_after_init;
 
@@ -1149,8 +1178,10 @@ static void __apply_fineibt(s32 *start_retpoline, s32 *end_retpoline,
 		goto err;
 
 	if (cfi_rand) {
-		if (builtin)
+		if (builtin) {
 			cfi_seed = get_random_u32();
+			cfi_bpf_hash = cfi_rehash(cfi_bpf_hash);
+		}
 
 		ret = cfi_rand_preamble(start_cfi, end_cfi);
 		if (ret)
diff --git a/arch/x86/net/bpf_jit_comp.c b/arch/x86/net/bpf_jit_comp.c
@@ -17,6 +17,7 @@
 #include <asm/nospec-branch.h>
 #include <asm/text-patching.h>
 #include <asm/unwind.h>
+#include <asm/cfi.h>
 
 static bool all_callee_regs_used[4] = {true, true, true, true};
 
@@ -51,9 +52,11 @@ static u8 *emit_code(u8 *ptr, u32 bytes, unsigned int len)
 	do { EMIT4(b1, b2, b3, b4); EMIT(off, 4); } while (0)
 
 #ifdef CONFIG_X86_KERNEL_IBT
-#define EMIT_ENDBR()	EMIT(gen_endbr(), 4)
+#define EMIT_ENDBR()		EMIT(gen_endbr(), 4)
+#define EMIT_ENDBR_POISON()	EMIT(gen_endbr_poison(), 4)
 #else
 #define EMIT_ENDBR()
+#define EMIT_ENDBR_POISON()
 #endif
 
 static bool is_imm8(int value)
@@ -304,6 +307,69 @@ static void pop_callee_regs(u8 **pprog, bool *callee_regs_used)
 	*pprog = prog;
 }
 
+/*
+ * Emit the various CFI preambles, see asm/cfi.h and the comments about FineIBT
+ * in arch/x86/kernel/alternative.c
+ */
+
+static void emit_fineibt(u8 **pprog)
+{
+	u8 *prog = *pprog;
+
+	EMIT_ENDBR();
+	EMIT3_off32(0x41, 0x81, 0xea, cfi_bpf_hash);	/* subl $hash, %r10d	*/
+	EMIT2(0x74, 0x07);				/* jz.d8 +7		*/
+	EMIT2(0x0f, 0x0b);				/* ud2			*/
+	EMIT1(0x90);					/* nop			*/
+	EMIT_ENDBR_POISON();
+
+	*pprog = prog;
+}
+
+static void emit_kcfi(u8 **pprog)
+{
+	u8 *prog = *pprog;
+
+	EMIT1_off32(0xb8, cfi_bpf_hash);		/* movl $hash, %eax	*/
+#ifdef CONFIG_CALL_PADDING
+	EMIT1(0x90);
+	EMIT1(0x90);
+	EMIT1(0x90);
+	EMIT1(0x90);
+	EMIT1(0x90);
+	EMIT1(0x90);
+	EMIT1(0x90);
+	EMIT1(0x90);
+	EMIT1(0x90);
+	EMIT1(0x90);
+	EMIT1(0x90);
+#endif
+	EMIT_ENDBR();
+
+	*pprog = prog;
+}
+
+static void emit_cfi(u8 **pprog)
+{
+	u8 *prog = *pprog;
+
+	switch (cfi_mode) {
+	case CFI_FINEIBT:
+		emit_fineibt(&prog);
+		break;
+
+	case CFI_KCFI:
+		emit_kcfi(&prog);
+		break;
+
+	default:
+		EMIT_ENDBR();
+		break;
+	}
+
+	*pprog = prog;
+}
+
 /*
  * Emit x86-64 prologue code for BPF program.
  * bpf_tail_call helper will skip the first X86_TAIL_CALL_OFFSET bytes
@@ -315,10 +381,10 @@ static void emit_prologue(u8 **pprog, u32 stack_depth, bool ebpf_from_cbpf,
 {
 	u8 *prog = *pprog;
 
+	emit_cfi(&prog);
 	/* BPF trampoline can be made to work without these nops,
 	 * but let's waste 5 bytes for now and optimize later
 	 */
-	EMIT_ENDBR();
 	memcpy(prog, x86_nops[5], X86_PATCH_SIZE);
 	prog += X86_PATCH_SIZE;
 	if (!ebpf_from_cbpf) {
@@ -3013,9 +3079,16 @@ struct bpf_prog *bpf_int_jit_compile(struct bpf_prog *prog)
 			jit_data->header = header;
 			jit_data->rw_header = rw_header;
 		}
-		prog->bpf_func = (void *)image;
+		/*
+		 * ctx.prog_offset is used when CFI preambles put code *before*
+		 * the function. See emit_cfi(). For FineIBT specifically this code
+		 * can also be executed and bpf_prog_kallsyms_add() will
+		 * generate an additional symbol to cover this, hence also
+		 * decrement proglen.
+		 */
+		prog->bpf_func = (void *)image + cfi_get_offset();
 		prog->jited = 1;
-		prog->jited_len = proglen;
+		prog->jited_len = proglen - cfi_get_offset();
 	} else {
 		prog = orig_prog;
 	}
@@ -3070,6 +3143,7 @@ void bpf_jit_free(struct bpf_prog *prog)
 			kvfree(jit_data->addrs);
 			kfree(jit_data);
 		}
+		prog->bpf_func = (void *)prog->bpf_func - cfi_get_offset();
 		hdr = bpf_jit_binary_pack_hdr(prog);
 		bpf_jit_binary_pack_free(hdr, NULL);
 		WARN_ON_ONCE(!bpf_prog_kallsyms_verify_off(prog));
diff --git a/include/linux/bpf.h b/include/linux/bpf.h
@@ -29,6 +29,7 @@
 #include <linux/rcupdate_trace.h>
 #include <linux/static_call.h>
 #include <linux/memcontrol.h>
+#include <linux/cfi.h>
 
 struct bpf_verifier_env;
 struct bpf_verifier_log;
@@ -1211,7 +1212,11 @@ struct bpf_dispatcher {
 #endif
 };
 
-static __always_inline __nocfi unsigned int bpf_dispatcher_nop_func(
+#ifndef __bpfcall
+#define __bpfcall __nocfi
+#endif
+
+static __always_inline __bpfcall unsigned int bpf_dispatcher_nop_func(
 	const void *ctx,
 	const struct bpf_insn *insnsi,
 	bpf_func_t bpf_func)
@@ -1303,7 +1308,7 @@ int arch_prepare_bpf_dispatcher(void *image, void *buf, s64 *funcs, int num_func
 
 #define DEFINE_BPF_DISPATCHER(name)					\
 	__BPF_DISPATCHER_SC(name);					\
-	noinline __nocfi unsigned int bpf_dispatcher_##name##_func(	\
+	noinline __bpfcall unsigned int bpf_dispatcher_##name##_func(	\
 		const void *ctx,					\
 		const struct bpf_insn *insnsi,				\
 		bpf_func_t bpf_func)					\
@@ -1453,6 +1458,9 @@ struct bpf_prog_aux {
 	struct bpf_kfunc_desc_tab *kfunc_tab;
 	struct bpf_kfunc_btf_tab *kfunc_btf_tab;
 	u32 size_poke_tab;
+#ifdef CONFIG_FINEIBT
+	struct bpf_ksym ksym_prefix;
+#endif
 	struct bpf_ksym ksym;
 	const struct bpf_prog_ops *ops;
 	struct bpf_map **used_maps;
diff --git a/include/linux/cfi.h b/include/linux/cfi.h
@@ -11,6 +11,13 @@
 #include <linux/module.h>
 #include <asm/cfi.h>
 
+#ifndef cfi_get_offset
+static inline int cfi_get_offset(void)
+{
+	return 0;
+}
+#endif
+
 #ifdef CONFIG_CFI_CLANG
 enum bug_trap_type report_cfi_failure(struct pt_regs *regs, unsigned long addr,
 				      unsigned long *target, u32 type);
diff --git a/kernel/bpf/core.c b/kernel/bpf/core.c
